reading DIP switches using few GPIO pins

[geekEE]

I'm trying to read a few 4-bit rotary DIP switches in a microcontroller in the most economical way possible. I'd like to minimize the number of GPIO pins used so that I don't have to buy a larger chip than absolutely necessary. I thought of two different ways and I wanted to see if anyone had any other tips or tricks to share.

1. I could use shift registers like the 74HC165 to read the switches and shift them in. This could be extended by cascading shift registers.

2. I could use resistors to encode the switches as analog levels and read them in using the 10-bit A/D in the microcontroller. I could extend this by using an analog mux to read in more switches. Does anyone know an arrangement of resistors that would maximize the resolution? The usual DAC R-2R tree doesn't work unless the switches are SPDT.

Anyone have any other tricks or comments?

 

[analogkid2digitalman]

8:1 ot 16:1 Multiplexer?

http://en.wikipedia.org/wiki/Multiplexer

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"It's the questions that drive us"
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 

[MacGyverS2000]

My answer was killed when I hit the wrong browser key and now the site is down, so it's not letting me submit my answer

Anyway, I think I have an answer using a modified resistor ladder:

+5V---+-S-+------+-S-+------+-S-+------+-S-+------+--Micro
      |-R-|      |-R-|      |-R-|      |-R-|      |
                                                  R
                                                  |
                                                 ===
                                                Ground

4 DIP switches in series, each with a parallel resistor, each resistor in the typical decade fashion.

An open switch includes the resistor in the divider, a closed switch bypasses it.

Had a blast coming up with this one!

Dan - Owner

http://www.Hi-TecDesigns.com

 

[MacGyverS2000]

Oh, rotary... I may have to rethink this one...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[itsmoked]

Bad day? LOL

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[logbook]

Dan, you get negative benefits using decade resistor values in your earlier posting. It eats up dynamic range and makes noise margin a problem. Equal resistors would have been fine had the OP not required a rotary 4 bit DIP switch.

 

[geekEE]

I see where you were going macgyvers, but I think it would still end up with the 16 different voltages being spread unevenly across the range between GND and VCC.

logbook, I don't think equal resistors would work there because then I would not be able to distinguish between the different switches being closed.

Of course, the point is moot because the 4-bit rotary DIP switches that I'm trying to read have four SPST switches with one common pin.

Thanks for everyone's input so far!

 

[logbook]

Sorry didn't mean to say equal resistors would work for your application, but as wired above they would have been better. The correct answer is

http://www.logbook.freeserve.co.uk/image/ladder.gif

Notice that these are the only resistors in the E24 standard range that give the exact 1-2-4 ratios!

 

[logbook]

output is from opamp and input line could be a negative voltage. The input can be digitally strobed. All inputs could be summed into the opamp avoiding an analog mux. Tolerances are left as an exercise for the designer.

 

[MacGyverS2000]

geek,

Uneven isn't a problem, as long as the smallest voltage change is within the resolution of your ADC (taing resistor tolerances into account). Still, it's a shame I didn't see the rotary part sooner... I'll think on it again when I get a moment at work today.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[jimkirk]

For simplicity, combine logbook's & macgyver's solution. Use positive reference for input line, remove op amp, and tie the right side of the 2.2K to ground.

Now you have a voltage divider with the output on the left end of the 2.2K, sent that to the ADC.

You'll have reduced dynamic range and linearity, but extracting 4 bits with a 10 bit ADC shouldn't be a problem.

You can look at this as approximating switchable current sources driving a resistive load. The smaller you make the load resistor (2.2K) compared to the binary weighted ones, the better your linearity, but the smaller your full scale.